« Previous | Contents | Next »
Listen
PUBLIC ATTITUDES TO THE ENVIRONMENT IN SCOTLAND - TECHNICAL REPORT
Chapter 2 - SAMPLING AND RESPONDENT SELECTION
2.1 DESIGN
The aim of the sample design was to obtain data which would be representative of the Scottish adult population (i.e. those who are 16+ years of age) who reside within households in Scotland
2. A probability sampling technique was employed with households being selected using a multi-stage stratified approach.
The sample was clustered in order to ensure that the fieldwork costs were constrained. However, to minimise the effect of clustering on the results, a large number of sample points (800) were specified with a maximum of eight interviews to be conducted at each point. The sample points were spread across the whole of Scotland, including the Highland and Islands. The intention was to achieve four thousand interviews from the initial selection of 6,400 households.
Within each sample point, two substitute addresses were also identified, to be used only if any of the original eight addresses proved to be invalid (see
chapter 3).
2.2 METHOD
The sampling frame was based on a cleaned version of the Postcode Address File (PAF) of residential addresses. Within the PAF, each of the Scottish postcodes were classified according to the rural/urban indicator used on the Scottish Household Survey, allocated to each by the Scottish Executive and updated by the sampling agency. The rural/urban indicators and their corresponding definitions are itemised in table 2.1 below.
Table 2.1
Definitions of rural/urban indicators
| Area type | Postcode units in |
1 | 'The primary cities' | Settlements of over 125,000 population (Aberdeen, Dundee, Glasgow and Edinburgh) |
2 | 'Other urban' | Other settlements over 10,000 population |
3 | 'Small accessible towns' | Settlements 3-10,000 population
and within a 30 minute drive time of a settlement of 10,000 or more |
4 | 'Small remote towns' | Settlements 3-10,000 population
and more than a 30 minute drive time of a settlement of 10,000 or more |
5 | 'Accessible rural' | Settlements of less than 3,000 population
and within a 30 minute drive time of a settlement of 10,000 or more |
6 | 'Remote rural' | Settlements of less than 3,000 population
and more than a 30 minute drive time of a settlement of 10,000 or more |
Each Scottish postcode was linked to an Enumeration District (ED) by using proximity analysis. From the postcode-to-ED link, it was possible to calculate how many addresses there were in each ED, as well as which was the predominant rural/urban category (based on postcodes within the ED). Using this methodology it was possible to allocate a rural/urban indicator to each Scottish ED. EDs that had less than 60 addresses within them were linked to the nearest large ED.
For stratification purposes, the rural/urban classification was taken as the predominant classifier. Stratification of the EDs was undertaken within the six rural/urban categories. It should be noted that the sample was not allocated in proportion to the number of households in each of the six rural/urban categories. Instead, the rural categories were deliberately over-represented and the urban categories under-represented to ensure adequate coverage of the more rural areas. The table below indicates the sampling fractions that were used for each rural/urban category. If boosting had not occurred the sampling fraction that would have been used is approximately 1:290. As the table indicates, addresses in the more urban areas were less likely to be selected for inclusion in the sample than those in more rural and remote areas.
Table 2.2
Sampling fractions used for each rural/urban area
Area type | Sampling fraction |
'The primary cities' | 1:343 |
'Other urban' | 1:343 |
'Small accessible towns' | 1:338 |
'Small remote towns' | 1:146 |
'Accessible rural' | 1:185 |
'Remote rural' | 1:194 |
The second level of stratification was undertaken within each rural/urban area according to the Mosaic code of individual EDs to ensure that the sample was, as far as possible, representative of Mosaic codes across Scotland. This meant sorting EDs by Scottish Mosaic within each of the six rural/urban categories. See
appendix 1 for a definition of Scottish Mosaic.
After stratification of the EDs, sample points were randomly selected and assigned to either sample A or sample B - this denoted which questionnaire would be administered at each sample point.
Then the profiles of the two samples were checked to ensure that there was a close match between them in terms of their Mosaic profile, the rural/urban indicator and by the ten broad geographic regions (identified at the sample design stage as being of possible analytical use subsequently). Discrepancies were highlighted and some reallocation of points introduced to improve the geographic match of the two samples. The total number of points swapped manually between sample A and sample B was 24, that is 12 points were taken from sample A and added to sample B and 12 others were moved from B to A.
2.3 OUTCOMES
Details of the final profiles of both sample A and sample B are provided in the following tables. Table 2.3 also shows the true population profile for residential addresses in Scotland, without adjustment.
Table 2.3
Comparison of sample A and sample B by rural/urban indicator
| Population Profile (residential addresses)
% | Sample A Design
% | Sample B Design
% |
The primary cities | 39.27 | 33.25 | 33.25 |
Other urban | 30.22 | 25.75 | 25.75 |
Small accessible towns | 9.53 | 8.25 | 8.25 |
Small remote towns | 2.95 | 6.00 | 6.00 |
Accessible rural | 11.85 | 18.00 | 18.00 |
Remote rural | 6.19 | 8.75 | 8.75 |
TOTAL | 100 | 100 | 100 |
The sample profile's geographic and mosaic profiles was also affected by the rural boosting.
The proportion of addresses in Aberdeen, Edinburgh, Glasgow and Dundee was less than that indicated by the population data while areas to the North and South of the central belt were boosted.
Table 2.4
Comparison of sample A and sample B by Local Authority groupings
Name | Population Profile
(residential addresses)
% | Sample A Design
% | Sample B Design
% |
Aberdeen | 4.45 | 4.25 | 4.25 |
Edinburgh | 9.36 | 8.50 | 7.75 |
Glasgow | 13.60 | 11.50 | 11.50 |
Dundee | 3.13 | 2.50 | 2.25 |
|
Highland area | 7.41 | 11.25 | 11.25 |
Tayside/Grampian area | 10.37 | 12.25 | 12.25 |
Central | 11.88 | 12.00 | 12.00 |
Lothian | 5.93 | 6.00 | 6.50 |
South | 4.97 | 5.25 | 5.75 |
Ayr | 7.04 | 6.25 | 6.25 |
Glasgow conurbation | 21.85 | 20.25 | 20.25 |
TOTAL | 100 | 100 | 100 |
A similar effect can be seen on the Mosaic profile - there is a slight under-representation of most groups, with the exception of those in the White Collar Owners group and the Open Countryside categories.
Table 2.5
Comparison of sample A and sample B by Mosaic code
Name | Population Profile
(residential addresses)
% | Sample A Design
% | Sample B Design
% |
Urban Establishment | 12.28 | 11.75 | 11.50 |
Burdened Borrowers | 10.61 | 10.00 | 10.00 |
Better Off Tenants | 14.63 | 14.50 | 13.75 |
Industrial Success | 7.34 | 7.00 | 7.25 |
Low Rise Council | 7.99 | 7.75 | 8.25 |
Council Flats | 6.01 | 5.25 | 5.25 |
Low Spending Elders | 7.30 | 6.50 | 6.50 |
Hi-Rise & Tenements | 6.25 | 5.25 | 5.50 |
Metro Lifestyles | 10.05 | 8.75 | 8.50 |
White Collar Owners | 13.01 | 16.75 | 16.75 |
Open Countryside | 4.41 | 6.50 | 6.75 |
TOTAL | 100 | 100 | 100 |
The main intent of the above tables is to demonstrate the extent to which a match was achieved in the sample design for the administration of the two versions of the questionnaire.
2.4 ADDRESS SELECTION
Within each sampling point a specified number of addresses were selected using the small users version of the Postal Address File. This is a listing of properties rather than residences and as such offers advantages and disadvantages when compared with the electoral register.
Firstly, a benefit of the PAF is that it is a more complete listing than the electoral register. The electoral register does not contain the names and addresses of everyone living in an ED. Indeed, an OPCS study estimated that almost 7% of the Scottish population had not registered to vote in 1991. Additionally, since 2001, legislation has specified that Local Authorities should maintain two versions of the electoral register, the full version and the edited version. The full version contains a list of all adults who are registered to vote in a Local Authorities area of remit. However, registered voters are entitled to opt out of the edited list. Therefore, if the edited electoral register were used as a sampling frame, not only would the frame be deficient of those who had not registered to vote, but it would also be deficient of those who had decided to opt out.
One of the acknowledged disadvantages of the PAF is that is contains a small number of properties which are not private residences, while it may also be inaccurate in other ways. However, procedures were put in place to make up for these deficiencies. The protocols developed for dealing with non-residential properties and other problems are described in chapter 3.
Another problem of using the PAF for sampling purposes is that the number of households at each property listed is not always correctly identified by the number of delivery points listed in the PAF. The most common inaccuracies are with respect to tenement properties and large blocks of flats. For these instances, interviewers were given a means by which they could randomly select an address themselves. The method by which interviewers could randomly select the addresses is described later within this chapter under 'unidentified multi-households'.
Standard address selection
To identify households where interviews were to take place, a random sample of addresses was drawn from the list of delivery points for each ED. Eight addresses (and two substitute addresses) were drawn for each sample point and issued to the interviewer. As mentioned already and discussed in detail in chapter 3, the substitute addresses were only to be used under specific circumstances.
Multiple households
Once the interviewer had ascertained that a specific address was residential, some way of working out how many households resided at an address was needed.
The definition of a household given to interviewers for this survey was:
'one person or a group of people who have the accommodation as their
only or main residence'
AND
(for a group of people) 'either share at least one main meal a day or share the living accommodation'.
Where an address contained only one household, this made the selection of possible respondents residing at an address straightforward. This was expected to be the case for the majority of addresses and the interviewer could go straight to the process of identifying the respondent (covered later within this chapter). The address selection processes for multi-household and unidentified multi-household addresses are covered next.
Multi-households
A multi-household address is one where there is more than one household at an address. For example, a house that is divided into self-contained bed-sits or if the address is simply listed as 24 High Street and it actually consisted of three flats, 24A, 24B and 24C.
PAF has been developed anticipating the existence of multi-household dwellings. The sample provided to interviewers contained details of the expected number of households at each address and also identified which household should be selected for interview, using a random procedure.
Where a multiple household was known to exist (in 419 cases representing 5% of the sample), this might show on the sample list as having a '3' for 'Number of households' and a '2' for 'Selected household'. This means that there were three households thought to be at the address and it was intended that the interviewer should conduct an interview at the second household. So if the household had flats designated 24A, 24B and 24C, the interview would be conducted at 24B.
Unidentified Multi-households
There remained the possibility that some multiple households might not be identified on PAF or that the details given might be wrong. Under these circumstances, a different protocol was required.
In these cases, the responsibility for selecting the appropriate address lay with the interviewer. To help interviewers select the appropriate address they were provided with a multi-household selection sheet.
If an address was identified as a multi-household address or found to be recorded incorrectly in the PAF, interviewers were required to complete the following tasks:
Systematically list all households at the address. Households were to be listed in some sort of order (i.e. if each flat had a number or a letter then the addresses would be listed in numerical or alphabetical order- flat 1, 2, 3 or flat A, B, C etc.) Otherwise, interviewers were instructed to start at the lowest floor and work in a clockwise direction, moving upwards, floor by floor, until either all households were listed or they had reached the 20
th line on the form.
Refer to the number in the column adjacent to the last household listed. (This gives the number of flats entered onto the list).
Read the number in the column adjacent to this. This number indicates which household to interview.
Contact and interview only at the selected household.
2.5 RESPONDENT SELECTION
Having established which households interviews would take place at, the next stage was to identify whom to select for interview.
All persons living at the address aged 16 years and over who resided at the address were listed. The following guidelines applied:
Students (at school, college or university) were only included if they were resident at that address at the time of the assignment.
Anyone who had been away from the address continuously for six months or more should be excluded even if the respondent continued to think of it as their main residence.
Men or women away from home in the Forces were not included.
Anyone who has been living continuously at an address for six months or more should be included at that address, even if their main residence was elsewhere.
Any adult should be included even if the address was a temporary one, for example, while they were searching for permanent accommodation. However, interviewers should not include anyone who was making a holiday or business visit only and who remains resident abroad.
People should not be included if the address was not their main residence (i.e., second or holiday home).
Once residency was established, all the residents were listed in alphabetical order from their first name. The respondent for the study was selected using a Kish Grid.
No individual within the household could be interviewed other than the selected individual.
2.6 FIELDWORK PROCEDURES
Each address within the main sample (excluding substitute addresses) was issued to interviewers on a Sample Issue Sheet (SIS - see
appendix 2 for an example). The SIS was completed to record details of contacts with particular households. Each SIS comprised the following components:
Address point identification details.
Household selection information
Respondent selection information
An outcome box for contact with households
An outcome box for contact with selected respondents
Questionnaire identifier - telling interviewer which questionnaire to use, version A or B.
2.7 substitution of addresses
As the PAF is not a perfect list, it was acknowledged that it might contain addresses where it was not possible or desirable to conduct an interview. An estimate was made of the likely level of invalid addresses that would be found in the original sample and the sampling company was asked to provide 2 addresses for each point, which could be used if any of the original 8 addresses were:
a business address
could not be located
empty
demolished
a non-main residence (second home)
an institution
The first of the substitute addresses to be used had to be the address closest to the address being substituted. If the first substitute address also turned out to be invalid and eligible for substitution, the substitution procedure could be repeated.
The same procedures for the identification of households and respondents were applied to substitute addresses as for the original addresses.
« Previous | Contents | Next »